Qualcomm technologies today announced the launch of Qualcomm's AI software stack portfolio to enhance Qualcomm's leading edge in the field of AI and intelligent networking. Qualcomm AI software stack is a complete set of AI solutions for OEMs and developers. It integrates and upgrades Qualcomm's various AI software products, and can support various intelligent terminals through rich AI software permissions and compatibility. Qualcomm said that this is the first time to realize the operation of enabling terminals across Qualcomm technologies through a single AI software portfolio, covering a wide range of smart Internet edge products, including smart phones, automobiles, XR, computing, the Internet of things and cloud platforms. Qualcomm AI software stack supports different AI frameworks and mainstream runtimes including tensorflow, pytorch and onnx, as well as developer libraries and services, system software, tools and compilers, so that any AI feature developed for a single terminal can be easily deployed on other terminals. Qualcomm said: "the complete Qualcomm AI software stack product portfolio is undoubtedly a great breakthrough. It will promote AI development by OEMs and developers, and bring high performance to Qualcomm technologies' broad product portfolio. We are pleased to further expand the company's unified technology roadmap and expand our leading edge at the edge of intelligent networking with this new product portfolio." This complete product portfolio supports direct access to the Qualcomm AI engine and the dedicated AI core of Qualcomm cloud AI 100 with Qualcomm AI engine direct. As part of the new solution, Qualcomm AI engine direct can now be extended on any AI accelerator in all Qualcomm technologies' products. As an AI library, it can directly entrust and deploy existing models to AI accelerators on Qualcomm technologies' platform. It can not only provide OEM manufacturers and developers with the function of model development, but also support the migration of newly developed models to different products and levels. The Qualcomm AI software stack portfolio also supports a range of tool suites, including Qualcomm AI model enhancement Toolkit (aimet), AI development graphical user interface (GUI), model analyzer for enhanced quantification and optimization, and neural network architecture search (NAS). At the 2021 Xiaolong technology summit, Qualcomm technologies and Google cloud announced the integration of Google cloud vertex AI NAS into the Qualcomm neural network processing SDK, enabling OEMs and ecosystems to create, test and deploy unique and efficient edge side experiences. The Qualcomm neural network processing SDK is still the key for OEMs and developers to run neural networks on Qualcomm technologies' products, so it will be upgraded and maintained regularly. The tool suite will provide the foundation for Qualcomm technologies to establish a unified SDK, so as to realize the deployment of cross intelligent networked terminals. (Xinhua News Agency)